Rho-Actin Signaling to the MRTF Coactivators Dominates the Immediate Transcriptional Response to Serum in Fibroblasts

Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

Rho-actin signaling to the MRTF coactivators dominates the immediate transcriptional response to serum in fibroblasts

Cyril Esnault,1 Aengus Stewart,2 Francesco Gualdrini,1 Phil East,2 Stuart Horswell,2 Nik Matthews,3 and Richard Treisman1,4 1Transcription Group, 2Bioinformatics and Biostatistics Group, 3Advanced Sequencing Facility, Cancer Research UK London Research Institute, London WC2A 3LY, United Kingdom

The transcription factor SRF (serum response factor) recruits two families of coactivators, the MRTFs (myocardin- related transcription factors) and the TCFs (ternary complex factors), to couple gene transcription to growth factor signaling. Here we investigated the role of the SRF network in the immediate transcriptional response of fibroblasts to serum stimulation. SRF recruited its cofactors in a gene-specific manner, and virtually all MRTF binding was directed by SRF. Much of SRF DNA binding was serum-inducible, reflecting a requirement for MRTF– SRF complex formation in nucleosome displacement. We identified 960 serum-responsive SRF target genes, which were mostly MRTF-controlled, as assessed by MRTF chromatin immunoprecipitation (ChIP) combined with deep sequencing (ChIP-seq) and/or sensitivity to MRTF-linked signals. MRTF activation facilitates RNA polymerase II (Pol II) recruitment or promoter escape according to gene context. MRTF targets encode regulators of the cytoskeleton, transcription, and cell growth, underpinning the role of SRF in cytoskeletal dynamics and mechanosensing. Finally, we show that specific activation of either MRTFs or TCFs can reset the circadian clock. [Keywords: SRF; MRTF; TCF; Rho; signal transduction; chromatin] Supplemental material is available for this article. Received February 3, 2014; revised version accepted March 20, 2014.

The SRF (serum response factor) transcription factor is an immediate transcriptional response and the roles played important regulator of cytoskeletal and muscle-specific by its cofactors have remained uncharacterized. gene expression (for reviews, see Posern and Treisman The myocardin and TCF family proteins compete for 2006; Olson and Nordheim 2010). SRF was first identified a common surface on the SRF DNA-binding domain but in studies of the fibroblast serum response, a classical also contact DNA flanking the SRF-binding site (Miralles model for cell cycle re-entry and wound healing. It et al. 2003; Wang et al. 2004; Zaromytidou et al. 2006). functions in partnership with members of two families Classic genomic footprinting studies suggest that SRF of signal-regulated cofactors: the MRTFs (myocardin-re- constitutively binds DNA and thereby defines the targets lated transcription factors; MRTF-A, MRTF-B, and myo- for its cofactors (Herrera et al. 1989). Functional studies cardin itself) and the TCF (ternary complex factor) family suggest that TCF and MRTF binding is gene-specific of Ets domain proteins (SAP-1, Elk-1, and Net). The (Gineitis and Treisman 2001; Miralles et al. 2003; Wang MRTFs, which bind G-actin, respond to fluctuations in et al. 2004), but the generality of this and its molecular G-actin concentration induced by Rho GTPase signaling basis remain to be determined. Where they have been (Miralles et al. 2003; Vartiainen et al. 2007), while TCF compared, the similarity of MRTF and SRF inactivation activity is controlled by Ras–ERK signaling. However, the phenotypes suggests that the MRTFs act solely through extent to which SRF is responsible for the serum-induced SRF (Medjkane et al. 2009; Mokalled et al. 2010). In contrast, the TCFs can act redundantly with other Ets proteins independently of SRF (for review, see Hollenhorst 4Corresponding author et al. 2011). Genomic studies revealed an association E-mail [email protected] Article published online ahead of print. Article and publication date are online at http://www.genesdev.org/cgi/doi/10.1101/gad.239327.114. Ó 2014 Esnault et al. This article, published in Genes & Development, Freely available online through the Genes & Development Open Access is available under a Creative Commons License (Attribution 4.0 In- option. ternational), as described at http://creativecommons.org/licenses/by/4.0.

GENES & DEVELOPMENT 28:000–000 Published by Cold Spring Harbor Laboratory Press; ISSN 0890-9369/14; www.genesdev.org 1 Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

Esnault et al. between SRF-binding sites and Ets, Sp1, AP-1, CREB, and stantially increased upon serum stimulation, reflecting NFY motifs (Valouev et al. 2008; Sullivan et al. 2011) and regulated MRTF nuclear accumulation; MRTF binding with CBP at signal-regulated enhancers (Kim et al. 2010). was sensitive to LatB but also exhibited a slight sensitiv- Here we revisit the role of the SRF network in the ity to U0126 (Fig. 2A,B; Supplemental Fig. S2D). MRTF-A fibroblast serum response and demonstrate a critical role and MRTF-B form homodimers (Supplemental Fig. S2E). for MRTF signaling. We identified >3100 SRF-binding Both were detected at many sites (Fig. 2C), and the sites, demonstrated that recruitment of SRF cofactors is correlation between MRTF-A and MRTF-B raw read indeed gene-specific, and showed that the majority of counts (Supplemental Fig. S2F) suggests that the appar- >2600 MRTF-binding sites exhibit MRTF-dependent SRF ently MRTF-B-specific peaks called in ChIP-seq reflect binding. We identified 960 serum-responsive SRF-linked poor MRTF-A antibody quality rather than MRTF-B- genes and showed that most are controlled through specific binding. Indeed, MRTF-A binding was detectable MRTF signaling, which activates transcription by pro- at such sites by ChIP coupled with quantitative PCR moting both RNA polymerase recruitment and promoter (ChIP-qPCR) (Supplemental Fig. S2G). In contrast, MRTF escape. We defined candidate gene sets controlled by the read counts correlated only poorly at the 75 apparently MRTF and TCF families. Our data suggest that MRTF– MRTF-A-specific sites, and these warrant further inves- SRF signaling is central to the serum response and tigation (Supplemental Fig. S2F). We also identified 121 strongly support the emerging idea that SRF signaling MRTF sites where SRF was not detectable (Fig. 2D; underlies the response of cells to mechanical cues. Supplemental Table S1), of which 57 were associated with transcribed sequences and/or RNA polymerase II Results (Pol II) at inducible or constitutive genes, while 64 were well-defined orphan binding sites (Supplemental Fig. S3). Much of SRF binding is controlled by Rho-actin Taken together, these data show that in NIH3T3 fibro- signaling blasts, the vast majority of MRTF sites are targeted by MRTF-A and MRTF-B homodimers or heterodimers, SRF represents the convergence point for serum-induced binding in association with SRF. Rho-actin and Ras–ERK signals (Fig. 1A). We used chromatin immunoprecipitation (ChIP) combined with deep sequencing (ChIP-seq) to define genomic binding sites in MRTF and TCF recruitment is SRF site-specific fibroblasts for SRF, the MRTFs, and the TCFs and cor- Inspection of the ChIP-seq data confirmed that at most related binding with changes in gene transcription 30 min binding sites, SRF exhibits a clear preference for either after serum stimulation, when SRF activation is maxi- the TCF or MRTF family cofactors (Fig. 1B). We identified mal. To assess the role of SRF-linked signal pathways 2215 discrete TCF peaks, of which ;10% colocalized in regulation, we used the pathway-specific inhibitors with SRF (Fig. 2D,E; Supplemental Fig. S2H; Supplemen- Latrunculin B (LatB) and U0126, which inhibit Rho-actin tal Table S1), a lower proportion than previously observed and ERK signaling, respectively, and the MRTF-specific for the Elk-1 TCF (Boros et al. 2009). At the majority of activator cytochalasin D (CD) (Supplemental Fig. S1). SRF–MRTF sites, TCF binding was undetectable even We established a core set of 3133 SRF-binding sites by under resting conditions, where MRTFs are cytoplasmic; ChIP-seq (see the Materials and Methods). In contrast to conversely, MRTF binding was undetectable at many expectations from classical studies of the c-fos gene SRF–TCF sites even under induced conditions, when (Herrera et al. 1989), the majority of SRF sites exhibited MRTFs are nuclear (Figs. 1B, 2D). Binding of both MRTFs an increased ChIP-seq signal upon serum stimulation and TCFs was detected at 123 SRF sites, but at 48 of these, (Fig. 1B). Using an inducibility threshold of 1.5, we one cofactor family or the other dominated (Fig. 2D,E). divided the SRF sites into 1000 ‘‘constitutive’’ and 2133 TCF-associated SRF binding was predominantly consti- ‘‘inducible’’ sites, with mean inducibility 1.03 6 0.01 and tutive (Supplemental Table S1). 2.51 6 0.04, respectively (Fig. 1C; see the Materials and These results clearly show that SRF cofactor recruit- Methods). LatB, but not U0126, specifically inhibited SRF ment is SRF site-specific even though the TCF ChIP-seq recruitment to inducible sites, suggesting that inducible data are of lower quality than the MRTF data. Using SRF binding reflects MRTF recruitment (Fig. 1D; Supple- HOMER (Heinz et al. 2010), we investigated whether mental Fig. S2A,B; see below). Almost 70% of SRF sites specific sequence motifs are associated with binding of were associated with the immediate 59 flanking region the different cofactors. The AP-1 and TEAD motifs were (À2 kb to the transcription start site [TSS]) or other gene preferentially associated with inducible, MRTF-associated features, and constitutive sites were selectively enriched sites, while SP-1, Ets, GFY, and NFY motifs were more within 2 kb of the TSS (Fig. 1E,F; Supplemental Table S1). strongly associated with constitutive, TCF-associated sites (Fig. 2F; see the Discussion). No cofactor binding MRTF binding is predominantly directed by SRF was detected at 755 SRF-binding sites (Fig. 2D; Supple- in fibroblasts mental Table S1). Motif analysis of these ‘‘solo’’ sites We defined 2416 MRTF-binding sites using ChIP-seq revealed all of the motifs associated with the MRTF- and (Figs. 1B, 2A; Supplemental Fig. S2C–G; Supplemental TCF-specific sites (Fig. 2F). Among the solo sites, 95 Table S1; see the Materials and Methods). MRTF binding exhibited LatB-sensitive serum-inducible SRF binding was low but detectable in unstimulated cells and sub- and were enriched in motifs associated with MRTF–SRF

2 GENES & DEVELOPMENT Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

MRTFs in serum-responsive transcription

Figure 1. Much of SRF binding is controlled by growth factor signaling. (A) Signaling pathways, activators, and inhibitors affecting the SRF network. (B) Representative SRF, MRTF, and TCF cofactor-binding profiles as normalized reads per base pair. Cell culture conditions were resting cells (0.3% FCS), cells stimulated for 30 min (15% FCS), stimulated in presence of LatB (LatB), and stimulated in presence of U0126 (U0126). (C) Serum stimulation induces SRF binding. Scatter plot comparing ChIP-seq read counts in stimulated and resting cells. (Dotted line) Linear regression plot for all sites (Spearman r, 0.3; fold-inducibility, 1.7 6 0.03). Note that division into inducible (>1.5-fold increase; red) and constitutive (<1.5-fold increase; black) populations greatly improves rank order correlations, respectively. Solid lines show linear regression plots for the two populations. (D) Metaprofile of SRF binding at constitutive and inducible sites. (E) SRF ChIP-seq peaks are associated with protein-coding genes (within 62 kb of the TSS, P < 10À999; within a gene feature, P = 1.9 3 10À106; basic x2 test). (Red) 59 flanking sequences (À2 kb to the TSS); (pink) other gene features (59 untranslated region [UTR], 338; introns, 740; coding exons, 39; 39 UTR, 11); (blue) intergenic, <70 kb from TSS or pA site; (gray) intergenic, >70 kb from TSS or pA site. (F) Distribution of SRF sites around the TSS, shown as sites per 40-base-pair (bp) bin. sites, while the remainder were enriched in motifs as- SRF–MRTF cooperate to exclude nucleosomes sociated with TCF–SRF sites (Supplemental Fig. S2I). at inducible binding sites Thus, although solo SRF sites may be targets of an as To ascertain the basis for inducible SRF binding, we yet unidentified SRF cofactor, it remains possible that examined the binding site sequences in more detail. they represent MRTF and TCF targets not detected by Almost 70% of SRF peaks contained perfect or singly ChIP-seq. More work is required to resolve this issue. mismatched copies of the SRF-binding CArG consensus

GENES & DEVELOPMENT 3 Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

Esnault et al.

Figure 2. Gene-specific MRTF and TCF recruitment. (A) Heat maps showing correlation of SRF and MRTF binding at inducible and constitutive SRF sites across a 2-kb region centered on the SRF peak summit. Color intensity represents normalized reads per 8-bp window. See Supplemental Figure S2, D and E. (B) Metaprofiles of MRTF binding, centered on SRF peak summits. (C) Most MRTF-A sites detected in ChIP-seq bind MRTF-B. See Supplemental Figure S2F. (D) Venn diagram showing relationships between sites bound by SRF, either MRTF or any TCF. See Supplemental Figure S2, G and H. (E) Scatter plot showing relative binding of MRTFs and TCFs by binding score (see the Supplemental Material). Among the 123 genes binding both families, more than twofold difference in score is associated with preferential response to ERK or Rho signaling (Gineitis and Treisman 2001). (F) SRF-associated sequence motifs within 100 bp of SRF peak summits, classified by cofactor-binding and SRF-binding inducibility. For motifs associated with LatB-sensitive and LatB-insensitive ‘‘no-cofactor’’ SRF sites, see Supplemental Figure S2I.

sequence CC(AT)6GG (almost 90%, including double the AP-1 and TEAD motifs were preferentially associ- mismatches) (Fig. 3A), with many containing multiple ated with strong CArG consensus matches, while the copies (Fig. 3B); ChIP-seq peak heights correlated with SP1, Ets, and GFY motifs were more prevalent at peaks the quality of the match to the CArG consensus (Fig. with a poor or no match to the CArG consensus (Fig. 3C). Surprisingly, inducible SRF-binding sites in general 3E). exhibited a better match to the CArG-box consensus We examined MRTF–SRF complex formation in vitro than constitutive ones, and, similarly, MRTF-associ- by electrophoretic mobility shift assay (EMSA) using ated SRF peaks matched the CArG consensus better probe sequences derived from inducible and constitutive than TCF-linked ones (Fig. 3D). Consistent with this, SRF sites (inducible: Ankrd1, Slc2a1, Trib1, and Prim2;

4 GENES & DEVELOPMENT Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

MRTFs in serum-responsive transcription

Figure 3. Characterization of MRTF- and TCF-associated SRF sites. (A) Match to the SRF CArG consensus (CCW6GG) within 100 bp of the SRF peak summit. Expected values are number of matches expected by chance with randomly selected sequences. (B) Multiple matches to the CArG consensus exist within 100 bp of each SRF peak summit. Peaks are grouped according to the best CArG match. (C) SRF peak height increases with matches to the CArG consensus. P < 0.0001, Mann-Whitney test. (D) Relationships between CArG consensus match and SRF-binding inducibility (left) or cofactor specificity (right). (E) SRF-associated sequence motif frequency plotted as function of match to the CArG consensus. (F) MRTF–SRF complex-binding properties at constitutive and inducible sites are similar in EMSA. (Left) EMSA analysis performed with whole-cell extract from NIH3T3 cells transfected with SRF expression plasmid (SRF) or vector alone (control) together with recombinant MRTF-A123-1A (non-actin-binding mutant) (Vartiainen et al. 2007). An asterisk marks the binding conditions used for antibody supershift assays (aSRF and aMRTF). (Right) Yield of the total MRTF–SRF complex quantified relative to SRF alone, taken as 1 (left plot) or percentage of maximum (right panel). Inducible and constitutive SRF sites are coded red and black, respectively. (G) H3 ChIP-seq metaprofiles at constitutive and inducible SRF sites under different assay conditions. (H) Evolutionary conservation across SRF-binding sites, determined by the Phastcons algorithm. (Left) All sites. (Right) Promoter- associated sites (transcription at the right). (I) Indirect cooperativity model for inducible SRF binding. Constitutive SRF binding and low nucleosome occupancy are facilitated by low nucleosome affinity and/or binding of other transcription factors nearby, and MRTF activation therefore has no effect. Inducible SRF binding is associated with high nucleosome affinity and/or the absence of other transcription factor-binding events; at these sites, SRF binding alone is insufficient for nucleosome displacement, which requires formation of the MRTF/SRF complex.

GENES & DEVELOPMENT 5 Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

Esnault et al. constitutive: Pdlim5, Ier2, and Fos). Each probe bound including 13 pre-microRNAs (pre-miRNAs), 11 small nu- SRF similarly in vitro even though SRF exhibited differ- cleolar RNAs (snoRNAs), one small nuclear RNA (snRNA), ential binding to these sites in vivo. Formation of the and 31 long intergenic noncoding RNA (lincRNA) clusters ternary MRTF-A–SRF complex resulted in a twofold to (Fig. 4B; Supplemental Fig. S5C; Supplemental Table S3). fourfold increase in DNA binding regardless of whether We next sought to define a maximum distance at which the complex was formed on a constitutive or inducible an SRF-binding site might be expected to influence probe (Fig. 3F). Thus, signal-inducible binding of SRF in transcription. SRF-binding sites were significantly more vivo cannot simply reflect cooperativity in the formation frequent within 70 kb of the 12,213 genes that were of the MRTF–SRF ternary complex. detectably transcribed in our cells (Fig. 4C; Supplemental We next investigated the relationship between SRF Table S4A) and also enriched within active gene features, binding and nucleosome displacement using H3 ChIP-seq even beyond 70 kb from the TSS (P < 10À4, Fisher’s test). and ENCODE DNase I sensitivity data (GSM1003831) (The Among the 1845 candidate SRF-regulated genes, we ENCODE Project Consortium 2011; Thurman et al. 2012). therefore considered only those with SRF sites within Both constitutive and inducible SRF-binding sites were 70 kb or within the gene feature as potential SRF targets, associated with a local minimum in the H3 ChIP signal, dividing them into two groups: ‘‘direct’’ genes with SRF which was more pronounced at constitutive SRF sites (Fig. sites within 2 kb 59 to the TSS or within a gene feature 3G; Supplemental Fig. S4A). Consistent with this, SRF- and ‘‘near’’ genes within 70 kb of an SRF site. binding sites colocalized with peaks of DNase I hypersen- Strikingly, of the 1976 direct genes, only 527 were sitivity (Thurman et al. 2012), with constitutive sites induced by SRF-linked signals, while expression of 1242 generally being more sensitive (Supplemental Fig. S4B,C). apparently constitutive genes remained unchanged (Sup- Serum stimulation significantly reduced H3 density at plemental Table S4A). As a consequence, only 658 (;32%) inducible but not constitutive SRF sites, and this was of the 2068 SRF-binding sites associated with direct genes blocked by LatB, indicating it required MRTF activation were associated with inducible transcription. However, (Fig. 3G; Supplemental Fig. S4A). These data suggest that the signal dependence of SRF binding and SRF cofactor inducible SRF binding reflects a requirement for MRTF– association was similar for both inducible and constitutive SRF complex formation in nucleosome displacement. genes (Fig. 4D; Supplemental Table S4B; see the Discus- We reasoned that constitutive SRF binding might re- sion). Inducible genes in the near class were frequently flect cooperation between SRF and other transcription associated with SRF sites that were closer to a second gene, factors in nucleosome displacement (Thurman et al. which could be inducibly or constitutively transcribed 2012) and so evaluated evolutionary conservation around (Fig. 4D,E; Supplemental Table S4B). Again, only ;30% constitutive and inducible sites. The DNA sequences (226/752) of the SRF sites 2–70 kb distant from their around constitutive SRF sites are evolutionarily con- nearest gene were associated with inducible transcription served across a substantially wider region than those (Fig. 4D; Supplemental Table S4B). SRF binding was pre- surrounding inducible SRF sites even when the latter dominantly associated with serum-induced gene activa- are located in promoter regions (Fig. 3H), which are tion. Only 67 repressed genes within 70 kb of an SRF site more evolutionarily conserved (The ENCODE Project appeared to be SRF-linked, compared with 960 that were Consortium 2007). Taken together, these data are consis- activated (Supplemental Table S2). tent with a model in which SRF acts with partner transcription factors to displace nucleosomes at constitutive MRTFs regulate the majority of serum-inducible genes binding sites and with MRTF to displace nucleosomes at The RNA-seq analysis above defines an SRF target gene set inducible binding sites (Fig. 3I; see the Discussion). of 960 serum-inducible genes (527 direct and 433 near). More than 95% of these (921 genes) could be classified as SRF binding is associated with both active candidate MRTF targets through MRTF ChIP, inhibition and serum-inducible genes by LatB, activation by CD, or more than one of these criteria (Fig. 4E; Supplemental Fig. S5; Supplemental Table We used RNA sequencing (RNA-seq) to determine how S4A), while a stringent MRTF–SRF-inducible gene set of transcription changes following a 30-min serum stimu- 683 genes was defined as those that both bound MRTFs lation, analyzing both total RNA-seq reads and intronic and were affected by the actin-binding drugs (Fig. 4E; RNA-seq reads to maximize our ability to detect gene Supplemental Table S5). Of the 33 SRF-linked ncRNA expression changes. We first identified 2144 serum- genes within 70 kb of an SRF site, all were candidate inducible genes (Fig. 4A, left). We then defined genes MRTF target genes (Supplemental Table S3). Finally, we potentially under SRF network control as those whose defined a further set of 76 serum-inducible genes as high- induction was significantly impaired by LatB and/or confidence TCF–SRF targets based on TCF binding and U0126 or that were also inducible by CD; this comparison sensitivity to U0126 (Supplemental Table S5). identified 1845 of these as candidate SRF-linked inducible genes at a fold discovery rate (FDR) of 0.08 (Fig. 4A, right; MRTF–SRF signaling promotes both Pol II recruitment Supplemental Table S2). A similar analysis identified 151 of and elongation 363 repressed genes identified as potential SRF-linked targets (Supplemental Table S2). In addition, we identified Having defined the SRF target gene set by RNA-seq, we 56 noncoding transcripts sensitive to SRF-linked signals, asked whether these genes exhibited changes in Pol II

6 GENES & DEVELOPMENT Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

MRTFs in serum-responsive transcription

Figure 4. The majority of serum-inducible genes are MRTF-controlled. (A, left) Scatter plot display of total (top) and intronic (bottom) RNA-seq read counts before and after serum stimulation. Serum-stimulated genes (FDR = 0.2) are highlighted in red. (Right) Definition of serum-inducible genes sensitive to SRF-linked signal pathways (FDR = 0.08). (Red bars) Median. See Supplemental Table S2. (B) Signaling to ncRNA genes, as in A. See Supplemental Table S3. (C) SRF sites are overrepresented within 70 kb of transcriptionally active genes. (Left) Frequencies of SRF sites relative to active and inactive genes (per 10-kb bin relative to TSS or pA site). Zero indicates sites within 2 kb of the TSS or within a gene feature. (Asterisks) Significant at P < 0.05, multiple t-test with Holmes-Sidak correction. (Right) SRF-binding sites are significantly closer to active genes. (Asterisks) Significant at P < 0.0001, Mann-Whitney test. See Supplemental Table S4. (D) Relationship between SRF peak properties and activity of their closest associated gene. The left bars indicate peaks classified according to activity of the closest gene, gray bars at the side indicate the fraction of these peaks associated with regulatory events at genes up to 70-kb distant, the center bars represent classification of peaks by SRF-binding inducibility, and the right bars represent classification of peaks by cofactor association. (Direct) Sites within 2 kb of 59 flanking sequences or within a gene feature; (near) sites within 70 kb of the TSS or pA site. See Supplemental Table S4. (E) Candidate MRTF serum-inducible genes (defined by MRTF binding, sensitivity to LatB or CD, or both) categorized by distance from the nearest SRF sites, as in D. Many inducible genes whose closest SRF site lies within 70 kb share that site with a second gene. See Supplemental Table S4A.

GENES & DEVELOPMENT 7 Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

Esnault et al. occupancy following serum stimulation using Pol II ChIP- Discussion). We exploited this observation to investigate at seq. Serum-responsive genes exhibited significantly in- which step MRTFs act in the transcription cycle by exam- creased Pol II loading following serum stimulation, the ining the effect of LatB on Pol II recruitment at 483 SRF increase being maximal for the direct genes (Fig. 5A; see the target genes whose serum induction was sensitive to LatB.

Figure 5. MRTF acts at Pol II recruitment and post-recruitment steps. (A) Serum-induced genes exhibit increased Pol II loading. RNA Pol II ChIP-seq analysis with antibodies as follows: 8WG16 (Pol II CTD un-P), total reads from À2kbto+1 kb from TSS; H14 (Pol II CTD S5P), total reads from À2 kb to the gene 39 end or +70 kb, the limit of Pol II progress after 30 min of stimulation; and H5 (Pol II CTD S2P), total reads from +1 to the gene 39 end or +70 kb. Statistical significance, Mann-Whitney test, (**) P < 0.01; (****) P < 0.0001. (B) Representative MRTF-B and Pol II ChIP-seq tracks on Acta2 and Klf7.(C) LatB inhibits serum induction of Klf7 and Acta2, assessed by qRT–PCR. SEM of three independent experiments. (D) MRTF is required for Pol II recruitment on a subset of target genes. Four- hundred-eighty-three genes were analyzed whose serum-induced activation was LatB-sensitive in the absence of U0126. (Left) Scatter plot summary of the Pol II ChIP-seq signal from À2to+1 kb around the TSS. Genes exhibiting >30% reduction (group I) are shown in black, and those exhibiting <30% reduction (group II) are in red. (Right) Summary of LatB’s effect on the two groups. Statistical significance, Mann-Whitney test, (***) P < 0.001. (E) LatB inhibits group I and group II gene RNA synthesis to a similar extent in RNA- seq. (F) Metaprofiles of Pol II ChIP-seq for group I and group II genes. (Top) 8WG16, H14, and H5 normalized ChIP-seq read counts are shown across gene loci, standardized to 20 kb, and flanking 5 kb. (Bottom) Read counts from À1kbto+1 kb from the TSS.

8 GENES & DEVELOPMENT Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

MRTFs in serum-responsive transcription

Examination of individual ChIP-seq profiles revealed Supplemental Table S8) even though YAP, which is also that at genes such as Acta2, Pol II recruitment to the a target for Rho signaling (Yu et al. 2012) appears to be vicinity of the TSS was substantially serum-inducible and constitutively active in our experimental conditions LatB-sensitive, while genes such as Klf7 exhibited sub- (Supplemental Fig. S1C–E). The MRTF–SRF signature stantial Pol II recruitment prior to MRTF activation (Fig. also includes many components of the Hippo signaling 5B). In both cases, however, production of mature RNA interactome (Fig. 6C; for references, see Supplemental was sensitive to LatB and therefore was MRTF-dependent Table S8). The TCF-dependent signature showed signifi- (Fig. 5C). We examined the effect of LatB on Pol II cant overlap with HeLa cell Elk1–SRF targets previously recruitment at the promoter region (À2 kb/+1 kb relative defined by ChIP–chip, TCF-dependent TCR-activated to the TSS) across the whole population. Setting a thresh- genes in thymocytes, and ERK-dependent genes activated old of >30% for the LatB-induced decrease in Pol II ChIP- by PDGF in fibroblasts or light in the suprachiasmatic seq reads allowed definition of two categories of MRTF nucleus (SCN) (Fig. 6C; Supplemental Table S8). target gene: Pol II recruitment to the 205 group I genes Constitutively active direct genes exhibited enrichment was LatB-sensitive, while Pol II recruitment to the 278 for genes involved in metabolism, DNA synthesis, gene group II genes was LatB-insensitive (Fig. 5D). Neverthe- expression, and cell growth (Fig. 6A; Supplemental Table S6) less, both groups exhibited similar sensitivity to LatB in and were functionally different from inducible genes, accord- RNA-seq analysis (Fig. 5E). Group I genes exhibited a ing to fold enrichment of individual GO terms (Wilcoxon greater degree of inducibility, as assessed by both intronic test P < 0.0001, for ‘‘direct’’ gene subsets). The failure of these RNA-seq reads and Pol II density. In both groups, Ser5 genes to respond to signals despite their association with and Ser2 phosphorylated Pol II C-terminal domain (CTD) SRF cofactors suggests that they are somehow refractory to ChIP-seq signals within the gene body were LatB-sensitive SRF-linked signals (see the Discussion). (Fig. 5F). Taken together, these data indicate that MRTF activation facilitates both Pol II recruitment and promoter Both MRTF- and TCF-linked signaling to SRF can escape according to gene context (see the Discussion). resynchronize the circadian clock Identification of components of the circadian clock among Ontology of MRTF–SRF and TCF–SRF target gene sets SRF target genes was intriguing given that serum shock can reset the circadian clock in fibroblasts (Balsalobre et al. Gene ontology (GO) analysis of the MRTF–SRF target 1998). Of SRF targets in the core clock network, Per1, Per2, genes using DAVID (Huang et al. 2009) revealed hundreds Nr1d1, Rora,andNfil3 bound MRTFs, but only Per1 of genes involved in actin filament dynamics, cell adhe- bound the TCFs (Fig. 7A,B; Supplemental Fig. S7A; for sion, extracellular matrix (ECM) synthesis and process- references, see Supplemental Table S8). Similar to serum ing, cell motility, and other actin-linked processes as well shock, MRTF activation by CD was sufficient to re- as a significant number of genes involved in microtubule- synchronize the fibroblast clock (Fig. 7C; Supplemental based cytoskeletal dynamics (Fig. 6A; Supplemental Fig. S7B,C), and this required SRF (Supplemental Fig. S7D). Fig. S6A; Supplemental Tables S6, S7). Consistent with We have shown elsewhere that systemic circadian activa- this, SRF inactivation was associated with defects in the tion of SRF target genes in the liver is associated with F-actin and microtubule cytoskeletons and increased circadian fluctuations in MRTF activity (Gerber et al. 2013). nuclear size and aspect ratio (Supplemental Fig. S6B). MAPK signaling induced by TPA was previously shown A second major MRTF–SRF target class includes tran- also to be sufficient to reset the fibroblast clock (Akashi scriptional machinery, chromatin regulators, and >70 and Nishida 2000). Since Per1 istheonlycoreclock transcription factors, including the classical AP1 and component that recruits TCFs, we tested whether it was Egr families and regulators of differentiation, cell mor- sufficient for clock resetting. In mouse embryonic fibro- phogenesis, and motility (Fig. 6A,B). The SRF target gene blasts (MEFs), TPA stimulation rapidly activated tPer1 bu set is also enriched in genes involved in cell growth and not Per2;however,Per2 and other core clock components metabolism, including circadian clock components, and were also activated 24 h later (Fig. 7D). Both the immediate genes controlled by systemic circadian cues (Fig. 6A,C; and longer-term transcriptional activation of clock tran- for references, see Supplemental Table S8). scription were abolished in MEFs lacking all three TCFs MRTF–SRF target genes overlap with gene signatures (Fig. 7D). Taken together, these data show that MAPK- associated with cancer cell invasiveness and metastasis, induced resetting of the circadian clock is mediated by response to ECM stiffness, or response to FAK or TGFb TCF–SRF signaling to Per1 in fibroblasts. Thus, Per1 and signaling (Fig. 6C,D; for references, see Supplemental Per2 represent SRF target genes linking the core clock Table S8). In several cases, the inducible MRTF–SRF target components to two different SRF-linked signal pathways, gene set exhibited significant overlap with gene sets Ras–ERK–TCF and Rho-actin–MRTF, which are triggered identified both as up-regulated or down-regulated between by different systemic and cell-specific cues (Fig. 7E). two experimental conditions. We feel this likely reflects the normalization procedures used for comparative analysis of microarray data sets, which generally assume Discussion similar overall RNA expression levels between conditions (for discussion, see Loven et al. 2012). The MRTF and In this study, we showed that Rho-actin signaling to the YAP–TAZ signatures also overlap significantly (Fig. 6C; MRTF family of SRF coactivators regulates multiple

GENES & DEVELOPMENT 9 Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

Figure 6. GO analysis of SRF targets. (A) SRF- and MRTF-linked gene signatures were analyzed using DAVID. (SRF) Serum-inducible genes with an SRF-binding site within 70 kb; (MRTF) serum-inducible genes with MRTF–SRF binding within 70 kb or LatB and/or CD sensitivity; (stringent MRTF) MRTF–SRF binding within 70 kb and sensitive to LatB or CD. The signatures are compared with SRF-linked serum- inducible or constitutively transcribed ‘‘direct’’ target genes (i.e., with SRF sites within 2 kb of a 59 flanking sequence or within a gene feature). See also Supplemental Tables S6 and S7. (F.E.) Fold enrichment. (B) SRF-controlled genes involved in transcriptional regulation are subdivided into functional categories. (C) Relationship between the inducible SRF, MRTF, and TCF gene signatures and previously defined sets of genes up-regulated between two experimental conditions; statistical significance by two-tailed Fisher test. See Supplemental Table S8. (D)MRTF– SRF signaling is a nuclear component of integrin-mediated ‘‘inside-out’’ signaling. Classes of SRF target genes involved in adhesion signaling and mechanosensation are shown. Engagement with the ECM induces changes in actin dynamics and actomyosin contractility, promoting focal adhesion assembly (blue arrows). MRTF–SRF target gene expression provides an additional long-term ‘‘inside-out’’ signaling mechanism (red arrows). The green dashed line indicates direct physical coupling between actomyosin, focal adhesions, and the nucleus. Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

MRTFs in serum-responsive transcription

Figure 7. SRF and MRTF can both reset the circadian clock. (A) SRF targets among circadian clock circuits (for discussion of regulatory loops, see Koike et al. 2012). (B) SRF, MRTF, and TCF ChIP-seq peaks are shown below in blue, red, and green, respectively, aligned with binding sites in the liver for the clock genes Per1 and Per2 (core loop components) and Nr1d1, Nr1d2, Dbp, and Nfil3 (interlocking loop components) (Koike et al. 2012), shown in gray. (C) Clock resetting by MRTF activation. NIH3T3 cells were treated with 2 mM CD for 2 h followed by washout; transcripts were quantified by qRT–PCR over 36 h. For serum stimulation kinetics, see Supplemental Figure S7C. (D) Clock resetting by TCF activation. Wild-type and SAP-1À/À Elk1À/À Net@/@ MEFs were treated with TPA, and transcripts were quantified by qRT–PCR. (E) Circadian clock synchronization by MRTF- and TCF-linked SRF signaling. aspects of the transcriptional response to serum stimula- that MRTF–SRF signaling constitutes a long-term aspect tion in fibroblasts. We identified ;3100 binding sites for of the ‘‘inside-out’’ cellular response to adhesion. Finally, SRF, associated with ;2600 genes. SRF acts as principal at the transcriptional level, MRTF–SRF signaling poten- gene targeting factor for the MRTFs, and much of SRF tiates both RNA Pol II recruitment and promoter escape binding is controlled through MRTF recruitment, which according to gene context. facilitates nucleosome displacement. The majority of 960 serum-inducible candidate SRF target genes that we SRF cofactor recruitment is gene-specific identified are regulated through MRTF–SRF signaling. We found that a substantial majority of the SRF sites in MRTF–SRF target genes include hundreds involved in fibroblasts recruit both MRTFs, with only a minority of actin cytoskeletal structures and dynamics, suggesting sites binding TCFs. In general, SRF sites exhibited a clear

GENES & DEVELOPMENT 11 Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

Esnault et al. preference for one cofactor family or the other, with only SRF to weak binding sites by cooperativity with MRTF. a small number of sites binding both at comparable Instead, we found that MRTF–SRF complex formation levels. Thus, SRF cofactor recruitment is gene-specific, induces nucleosome displacement at inducible SRF- consistent with predictions from functional studies of binding sites but that constitutive SRF sites remain model genes (Gineitis and Treisman 2001) and biochem- unaffected. ical experiments (Miralles et al. 2003; Wang et al. 2004; We propose that constitutive SRF binding reflects Zaromytidou et al. 2006). MRTF-A and MRTF-B form cooperation with transcription factors binding nearby to heterodimers, and both were detectable at the majority of induce nucleosome displacement even under resting con- MRTF sites. SRF appears to be the primary targeting ditions. Previous studies have shown that nucleosome agent for the MRTFs in fibroblasts: Ninety-five percent of displacement is facilitated by multiple independent tran- 2416 MRTF-binding events detected in fibroblasts were scription factor-binding events (Boyes and Felsenfeld SRF-associated, and we found no evidence for MRTF 1996), while at transcription factor clusters, aggregate recruitment by other sequence-specific DNA-binding pro- ChIP-seq peak heights correlate with DNase I sensitivity teins as proposed by others (Qiu et al. 2005; Morita et al. (Thurman et al. 2012). Consistent with this, constitutive 2007). The functional significance of the small number SRF sites are embedded in broad regions of evolutionary of apparently SRF-independent MRTF–DNA interactions conservation with a greater frequency of associated binding and a small number of apparently MRTF-A-specific motifs. In contrast, inducible sites are found in narrower binding events remains unclear; these include associa- regions of conservation, suggesting that cooperating tran- tions with transcribed sequences themselves and a small scription factors are less abundant and that it is MRTF number of intergenic sequences. Cofactor binding could recruitment that increases the probability of nucleosome not be detected at ;25% of SRF sites. At these sites, SRF displacement (Fig. 3I). might act with hitherto undetected cofactors or perhaps constitutively activate transcription through its C-terminal SRF is associated with transcriptionally active genes transcription activation domain (Johansen and Prywes 1993); however, examination of the sequence motifs asso- SRF-binding sites are preferentially associated with tran- ciated with them suggests that they might also represent scriptionally active genes, the majority being located undetected MRTF- or TCF-associated sites (see below). within gene features. We exploited this and the known Our data demonstrate cofactor-specific association be- involvement of SRF with serum-regulated transcription to tween SRF binding and other transcription factor motifs. define an SRF target gene set comprising the 960 serum- The AP-1 or TEAD motifs are frequently found at MRTF- inducible genes within 70 kb of an SRF site, almost 90% specific SRF sites, and it is conceivable that the MRTFs of which were dependent on SRF cofactor-linked signal interact directly with these: The MRTFs are known to pathways for transcription. Among these, we identified make DNA contacts, and their SAP domains have been an MRTF–SRF target gene set of 921 genes, as assessed by implicated in promoter selectivity (Wang et al. 2001; MRTF ChIP and/or sensitivity to MRTF-linked signals, and Zaromytidou et al. 2006). Alternatively, the TEAD and a stringent set of 683 MRTF–SRF targets satisfying both AP-1 motifs may reflect functional cooperation between criteria. We also identified a 76-gene TCF–SRF-inducible MRTFs and their cognate transcription factors, and, in gene signature based on TCF binding and U0126 sensitiv- this respect, the TEAD motifs are particularly intrigu- ity. Analysis of Pol II loading on serum-inducible genes ing given the apparent overlap between YAP–TAZ and showed a steady decrease of distance from the nearest SRF MRTF–SRF target genes (see below). In contrast, we found site, suggesting that at the time point that we analyzed, the that the Ets, NFY, and SP1 motifs, previously demon- primary determinant of transcription rate is SRF activity. strated to be SRF-associated (Valouev et al. 2008; Sullivan Our data show that SRF is overwhelmingly associated et al. 2011), were associated specifically with TCF–SRF- with serum-induced transcriptional activation. Although binding events. As discussed below, the enrichment of >14 times as many genes were activated as were re- these motifs at constitutive SRF sites with poor matches pressed, the molecular basis of repression should be to the CArG-binding consensus may reflect cooperativity studied further. More significantly, we found that more between SRF and their cognate factors in nucleosome than two-thirds of the genes associated with SRF sites displacement. are apparently constitutively transcribed. These genes, which are ontologically distinct from the serum-inducible SRF target set, also bind SRF cofactors but show no Most MRTF–SRF binding is signal-regulated response to serum stimulation even though the SRF A substantial majority of SRF-binding events are poten- network is maximally active at the 30-min time point tiated by MRTF activation, a surprising finding given analyzed. Indeed, examination of SRF targets expressed in classical genomic footprinting studies (Herrera et al. both NIH3T3 cells and MEFs revealed that only 8% of the 1989), although inducible binding has been reported more constitutively active SRF-linked genes were dependent recently (Kim et al. 2010; Leitner et al. 2011). Our in vitro on SRF, compared with 56% of the shared serum-in- MRTF–SRF complex formation experiments and obser- ducible genes (C Esnault and R Treisman, unpubl.). Pro- vation that inducible SRF sites are better matched to the moter-proximal SRF binding was previously observed in binding consensus than constitutive ones show that in- macrophages and, similarly, expression of only 6% of ducible SRF binding cannot simply reflect recruitment of these genes was dependent on SRF (Sullivan et al. 2011). It

12 GENES & DEVELOPMENT Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

MRTFs in serum-responsive transcription is likely that constitutive SRF-linked promoters are laps with that of light-induced gene transcription in the somehow refractory to transcriptional activation, although SCN, which induces Per1 but not Per2 transcription. we cannot exclude the possibility that they become active at late times, and it will be interesting to determine the MRTF–SRF signaling and mechanosensing molecular basis for this. The MRTF–SRF gene signature significantly overlaps with gene signatures characteristic of cancer cell invasion MRTF and transcription activation and metastasis and the response to mechanical stress. Members of the TCF family of SRF cofactors interact Integrin engagement with the ECM leads to FA assembly, with MED23 and other mediator subunits to promote Pol requiring Rho-dependent actomyosin contractility, and II recruitment and promoter escape at TCF–SRF target this process, known as inside-out signaling, is strongly genes such as Egr1 (for references, see Balamotis et al. implicated in cancer cell invasiveness (Paszek et al. 2005; 2009). Our RNA Pol II ChIP-seq data indicate that while for overview, see Butcher et al. 2009). Our data show that active MRTF is required for both RNA Pol II recruitment MRTF–SRF pathway activation constitutes an additional and activation at about half of its targets, it is required for long-term nuclear arm of inside-out signaling, controlling only a post-recruitment step at the others. Thus, in expression of the major FA force sensors p130Cas/Bcar1, addition to promoting Pol II recruitment, the MRTFs also Talin/Tln1, and Vinculin/Vcl and many actomyosin com- act at a post-recruitment step in transcriptional activa- ponents and regulators (Fig. 6D; Supplemental Table S7). tion. We previously reported that confinement of MRTFs An additional aspect of MRTF–SRF inside-out signaling is to the nucleus in resting cells does not activate transcrip- its potential effects on the ECM itself, since it controls tion (Vartiainen et al. 2007) and are currently investigat- expression of collagens, their modifiers (Lox and Plod3), ing the relationship between actin–MRTF interaction and and matrix metalloproteases (mmp9 and serpine1) (Fig. the different steps of transcription activation. 6D; Supplemental Table S7). Indeed, Lox expression promotes FAK-dependent matrix stiffening, FA assembly, Functional significance of MRTF–SRF and TCF–SRF and invasiveness (for review, see Barker et al. 2012). target gene sets Matrix stiffness is an important determinant of mechanical stress that induces MRTF activation (Huang Our results reveal that in fibroblasts, it is MRTF–SRF et al. 2012), and the MRTF–SRF signature overlaps signaling that underlies the role of SRF in cytoskeletal significantly with the matrix stiffness-associated breast dynamics. Within the 921-gene MRTF–SRF signature, cancer invasiveness and FAK-dependent gene signatures gene targets include components of both the actin and mi- (Provenzano et al. 2008, 2009). Matrix stiffness enhances crotubule cytoskeletons, cell–cell and cell matrix junctions, osteogenic differentiation while inhibiting adipogenesis ECM components, and vesicle trafficking components. (for review, see Butcher et al. 2009; Discher et al. 2009); The involvement of Rho-actin signaling in microtubule- consistent with this, Rho signaling to SRF promotes oste- associated gene expression is perhaps not surprising given ogenesis and inhibits adipogenic differentiation (Sordella theroleofRhoGTPasesignalingintheresponseto et al. 2003; Mikkelsen et al. 2010; Chen et al. 2012). Matrix challenge to microtubule integrity (for discussion, see stiffening also induces increased expression of Lmna,an Krendel et al. 2002). The MRTF-dependent serum re- SRF target, which is required for effective differentiation sponse also includes numerous transcriptional regula- along muscle and bone lineages (Swift et al. 2013) and tory factors and a significant number of genes involved stiffening of the nucleus itself (Lammerding et al. 2006). in cell growth and metabolism, including many of the Interestingly, Lmna inactivation results in increased components of the core circadian clock network. G-actin levels and decreased MRTF activity, probably SRF in circadian clock regulation through deregulation of emerin (Ho et al. 2013). The relationship between matrix stiffness, G-actin concentration, Previous work has shown that in fibroblasts, both serum and MRTF activity deserves systematic investigation. stimulation and ERK activation can reset the circadian clock (Balsalobre et al. 1998; Akashi and Nishida 2000). SRF signaling and the YAP–TAZ pathway Our results show that the core clock circuit genes are SRF targets and that both MRTF- and TCF-linked signals can The YAP–TAZ pathway is also mechanoresponsive but reset the clock. We have shown elsewhere that systemic is thought to be independent of MRTF–SRF signaling circadian cues promote oscillations of G-actin concentra- (Dupont et al. 2011; Calvo et al. 2013). Although in our tion that control MRTF–SRF signaling in the liver (Gerber experiments YAP is nuclear and presumably active, our et al. 2013). Here we showed that in fibroblasts, only Per1 MRTF–SRF signature overlaps significantly with those is a specific TCF–SRF target gene, and ERK activation published for YAP and includes Ctgf, Cyr61, and Ankrd1, induces an immediate transcriptional response of Per1, genes frequently used as readouts for YAP activation. resynchronizing the core clock in a TCF-dependent man- Moreover, MRTF–SRF signaling may also influence YAP ner. MAPK signaling has been previously shown to induce activity indirectly, as the MRTF signature includes YAP– immediate–early gene expression and reset the clock in TAZ pathway and Hippo interactome components (Kwon the SCN (Obrietan et al. 1998). Our data suggest that SCN et al. 2013), including TAZ/Wwtr1, Tead1, and Runx2 clock resetting will reflect TCF-dependent Per1 activation: (Fig. 6C; Supplemental Table S8). These observations and Indeed, the TCF-dependent gene signature strongly over- the occurrence of the TEAD motif at MRTF–SRF sites

GENES & DEVELOPMENT 13 Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

Esnault et al. suggest that the relationship between YAP/TAZ and threshold of P < 0.05. Constitutive and inducible SRF sites were MRTF signaling should be investigated further. defined using an inducibility threshold (SRF signal in 15% FCS/ signal in 0.3% FCS) such that the linear regression curve was SRF signaling and ncRNAs closest to 1 for the constitutive SRF population, which by definition should not be influenced by signal. We identified 33 serum-inducible ncRNAs as SRF targets, including miRNAs, snoRNAs, and long ncRNAs (lncRNAs). MRTF-A and MRTF-B Single ChIPs from resting cells and serum- Where known, the properties of these ncRNAs suggest stimulated cells with or without LatB or U0126 were analyzed. that they may also contribute to the serum response Two-hundred-seventy-four MRTF-A and 1178 MRTF-B peaks P À5 (Supplemental Fig. S5C). For example, miR-143 and miR- were called by MACS at < 10 and detected (1) in more than one sample for each antibody or (2) in the same condition by both 145 affect cytoskeletal dynamics in SMCs, miR-199a2 MRTF-A and MRTF-B antibodies, since MRTF-A and MRTF-B and miR-214 control myoblast differentiation, and miR- heterodimerize (Supplemental Fig. S2E). A further 1121 MRTF-A 21 promotes fibrosis and epithelial-to-mesenchymal and 1165 MRTF-B peaks were called as coincident with an transition (EMT) and is implicated in cancer, while inducible, LatB-sensitive SRF site at MACS P < 0.05, since SRF miR-22 is implicated in regulation of SRF itself in and MRTF read counts correlate well at inducible SRF peaks a cardiac overload model. The lncRNAs controlled by (Supplemental Fig. S2C). MRTF-A and MRTF-B read counts MRTF–SRF signaling include Tug1 and Malat1/Neat2 strongly correlate at sites called for both proteins (Supplemental (a lncRNA implicated in lung cancer cell invasion and Fig. S2F, left) and those called specifically for MRTF-B (Supple- metastasis), which are recruited by unmethylated and mental Fig. S2F, center), suggesting that failure to detect MRTF-A- methylated PC2, respectively. Intriguingly, PC2 itself and MRTF-B-specific sites generally reflects MRTF-A antibody quality; indeed, MRTF-A binding could be detected at such sites and KDM4C, the PC2 demethylase, are also MRTF by conventional ChIP-qPCR (Supplemental Fig. S2G). targets. Further work will be required to establish their function during serum stimulation. Finally, it will be in- TCFs—SAP-1, Elk-1, and Net We analyzed a single ChIP from teresting to investigate whether SRF-dependent lncRNAs resting cells or serum-stimulated cells with or without LatB or such as GM15270/Ctgf and GM13270/Rsu1 are in- U0126. Peaks were called at MACS P < 10À5 as those exhibiting volved in the regulation of SRF-controlled protein- signal in (1) more than one condition for the same family member coding genes in their vicinity. or (2) for multiple family members in the same condition.

RNA Pol II Pol II signal was quantified for all RefSeq genes and Concluding remarks their promoters. Enrichment was determined counting the num- MRTF–SRF signaling is an important aspect of the re- ber of normalized reads in 500-base-pair (bp) windows from À2kb sponse to growth factor signaling and an important medi- to +70 kb from the TSS or to the gene end if shorter. Metaprofiles ator of the cytoskeletal response to Rho GTPase activa- were generated as nucleotide average read density for the Pol II data sets across gene loci, standardized to 20 kb (65kb). tion. The MRTF–SRF will provide an interesting test bed for investigation of the relationships between cell–cell H3 Read counts were normalized to a total of 300 million reads, and cell–substrate interactions, cytoskeletal dynamics, and and the normalized read density per base pair was calculated. gene expression. Metaprofiles were average read density profiles centered on the SRF peak summits.

Materials and Methods RNA Total RNA was isolated using GenElute kit (Sigma). For qPCR ChIP and ChIP-seq quantitation, 500 ng of RNA was subjected to cDNA synthesis ChIP was performed as described (Miralles et al. 2003; Costello using the SuperScript III first strand synthesis system and et al. 2010) with minor modifications. Additional antibodies used random hexamer primers (Invitrogen). Data are from at least were MRTF-A (sc-21558), MRTF-B (sc-47282), Pol II CTD S2unP, three independent experiments. Libraries were prepared using and 8WG16 (sc-56767), all from Santa Cruz Biotechnology; total the Directional mRNA-seq Library Preparation version 1.0 pre- H3 (ab1791, Abcam); Pol II CTD S2P (H5, Covance); and Pol II release protocol (Illumina) with minor adjustments. To mini- CTD S5P (H14, Covance). See the Supplemental Material. mize the ribosomal content, we used DSN (Evrogen JSC) or Ribozero (Epicentre) treatment. ChIP-seq peak calling strategy RNA-seq SRF sites Two independent chromatin preparations from resting and serum-stimulated cells and single preparations from cells After RNA library preparation and sequencing, we screened the stimulated in the presence of inhibitors were used for ChIP-seq raw data against protein-coding gene and ncRNA gene databases with a pooled bead-alone control. Peaks were called using (RefSeq [release 47] and Ensembl [release 69], respectively) and a stringent MACS threshold of P < 10À5 (Zhang et al. 2008). derived data sets comprising either total read count across a given A core set of 3040 sites was defined as sites detected in both gene (‘‘all reads’’) or within introns only (‘‘intronic reads’’). Ex- resting samples + sites detected in both stimulated samples + pression levels in unstimulated cells were fitted to a Gaussian, sites detected in a LatB sample and any other sample + sites and 6664 genes were selected whose expression following stimu- detected in a U0126 sample and any other sample. An additional lation differed by no more than 1 SD from the mean in unstimu- 93 SRF-binding sites were defined on the basis that they co- lated cells. The mean expression level of these 6664 genes was incided with MRTF peaks called by MACS at P < 10À5, were then used to normalize read counts between different experimen- serum-inducible and LatB-sensitive, and passed a low MACS tal conditions. Differential expression analysis was performed

14 GENES & DEVELOPMENT Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

MRTFs in serum-responsive transcription with DESeq (Anders and Huber 2010) at an estimated FDR of Discher DE, Mooney DJ, Zandstra PW. 2009. Growth factors, <0.08. The serum-stimulated gene set was defined as those genes matrices, and forces combine and control stem cells. Science whose normalized ‘‘all reads’’ and/or ‘‘intronic reads’’ expression 324: 1673–1677. changes significantly. For details, see the Supplemental Material. Dupont S, Morsut L, Aragona M, Enzo E, Giulitti S, Cordenonsi M, Zanconato F, Le Digabel J, Forcato M, Bicciato S, et al. 2011. Role of YAP/TAZ in mechanotransduction. Nature 474: 179–183. Data access The ENCODE Project Consortium. 2007. Identification and ChIP-seq and RNA-seq data are available under Gene Expression analysis of functional elements in 1% of the human genome Omnibus accession number GSE45888. by the ENCODE pilot project. Nature 447: 799–816. The ENCODE Project Consortium. 2011. A user’s guide to the encyclopedia of DNA elements (ENCODE). PLoS Biol 9: Acknowledgments e1001046. We thank Rob Nicolas and Souhila Medjkane for help and advice Gerber A, Esnault C, Aubert G, Treisman R, Pralong F, Schibler during the early stages of this project; Benjamin Philimore and U. 2013. Blood-borne circadian signal stimulates daily oscil- Sharmin Begum for library preparation; and Caroline Hill, Nick lations in actin dynamics and SRF activity. Cell 152: 492– Luscombe, Michael Way, and members of the Transcription 503. Group for helpful discussions and comments on the manuscript. Gineitis D, Treisman R. 2001. Differential usage of signal This work was funded by Cancer Research UK core funding to transduction pathways defines two types of serum response the London Research Institute and Advanced Grant from the factor target gene. J Biol Chem 276: 24531–24539. European Research Council (ERC) to R.T. (Project 268690). C.E. Heinz S, Benner C, Spann N, Bertolino E, Lin YC, Laslo P, Cheng was initially supported by an EMBO long-term fellowship, and JX, Murre C, Singh H, Glass CK. 2010. Simple combinations C.E. and F.G. were partly supported by ERC to R.T. (Project of lineage-determining transcription factors prime cis-regu- 268690). C.E. and F.G. performed experiments and analyzed data. latory elements required for macrophage and B cell identi- R.T. and C.E. conceived the study, designed experiments, and ties. Mol Cell 38: 576–589. wrote the manuscript. N.M. prepared libraries and generated Herrera RE, Shaw PE, Nordheim A. 1989. Occupation of the sequence data. A.S., P.E., and S.H. analyzed data. c-fos serum response element in vivo by a multi-protein complex is unaltered by growth factor induction. Nature 340: 68–70. References Ho CY, Jaalouk DE, Vartiainen MK, Lammerding J. 2013. Lamin A/C and emerin regulate MKL1-SRF activity by modulating Akashi M, Nishida E. 2000. Involvement of the MAP kinase actin dynamics. Nature 497: 507–511. cascade in resetting of the mammalian circadian clock. Genes Hollenhorst PC, McIntosh LP, Graves BJ. 2011. Genomic and Dev 14: 645–649. biochemical insights into the specificity of ETS transcription Anders S, Huber W. 2010. Differential expression analysis for factors. Annu Rev Biochem 80: 437–471. sequence count data. Genome Biol 11: R106. Huang DW, Sherman BT, Lempicki RA. 2009. Systematic and Balamotis MA, Pennella MA, Stevens JL, Wasylyk B, Belmont integrative analysis of large gene lists using DAVID bioin- AS, Berk AJ. 2009. Complexity in transcription control at formatics resources. Nat Protoc 4: 44–57. the activation domain-mediator interface. Sci Signal 2: Huang X, Yang N, Fiore VF, Barker TH, Sun Y, Morris SW, Ding ra20. Q, Thannickal VJ, Zhou Y. 2012. Matrix stiffness-induced Balsalobre A, Damiola F, Schibler U. 1998. A serum shock in- myofibroblast differentiation is mediated by intrinsic mecha- duces circadian gene expression in mammalian tissue cul- notransduction. Am J Respir Cell Mol Biol 47: 340–348. ture cells. Cell 93: 929–937. Johansen FE, Prywes R. 1993. Identification of transcriptional Barker HE, Cox TR, Erler JT. 2012. The rationale for targeting activation and inhibitory domains in serum response factor the LOX family in cancer. Nat Rev Cancer 12: 540–552. (SRF) by using GAL4-SRF constructs. Mol Cell Biol 13: 4640– Boros J, Donaldson IJ, O’Donnell A, Odrowaz ZA, Zeef L, Lupien 4647. M, Meyer CA, Liu XS, Brown M, Sharrocks AD. 2009. Kim TK, Hemberg M, Gray JM, Costa AM, Bear DM, Wu J, Elucidation of the ELK1 target gene network reveals a role Harmin DA, Laptewicz M, Barbara-Haley K, Kuersten S, in the coordinate regulation of core components of the gene et al. 2010. Widespread transcription at neuronal activity- regulation machinery. Genome Res 19: 1963–1973. regulated enhancers. Nature 465: 182–187. Boyes J, Felsenfeld G. 1996. Tissue-specific factors additively Koike N, Yoo SH, Huang HC, Kumar V, Lee C, Kim TK, increase the probability of the all-or-none formation of Takahashi JS. 2012. Transcriptional architecture and chro- a hypersensitive site. EMBO J 15: 2496–2507. matin landscape of the core circadian clock in mammals. Butcher DT, Alliston T, Weaver VM. 2009. A tense situation: Science 338: 349–354. forcing tumour progression. Nat Rev Cancer 9: 108–122. Krendel M, Zenke FT, Bokoch GM. 2002. Nucleotide exchange Calvo F, Ege N, Grande-Garcia A, Hooper S, Jenkins RP, factor GEF-H1 mediates cross-talk between microtubules Chaudhry SI, Harrington K, Williamson P, Moeendarbary E, and the actin cytoskeleton. Nat Cell Biol 4: 294–301. Charras G, et al. 2013. Mechanotransduction and YAP-dependent Kwon Y, Vinayagam A, Sun X, Dephoure N, Gygi SP, Hong P, matrix remodelling is required for the generation and mainte- Perrimon N. 2013. The Hippo signaling pathway inter- nance of cancer-associated fibroblasts. Nat Cell Biol 15: 637–646. actome. Science 342: 737–740. Chen J, Yuan K, Mao X, Miano JM, Wu H, Chen Y. 2012. Serum Lammerding J, Fong LG, Ji JY, Reue K, Stewart CL, Young SG, response factor regulates bone formation via IGF-1 and Lee RT. 2006. Lamins A and C but not lamin B1 regulate Runx2 signals. J Bone Miner Res 27: 1659–1668. nuclear mechanics. J Biol Chem 281: 25768–25780. Costello P, Nicolas R, Willoughby J, Wasylyk B, Nordheim A, Leitner L, Shaposhnikov D, Mengel A, Descot A, Julien S, Treisman R. 2010. Ternary complex factors SAP-1 and Elk-1, Hoffmann R, Posern G. 2011. MAL/MRTF-A controls mi- but not net, are functionally equivalent in thymocyte de- gration of non-invasive cells by upregulation of cytoskeleton- velopment. J Immunol 185: 1082–1092. associated proteins. J Cell Sci 124: 4318–4331.

GENES & DEVELOPMENT 15 Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

Esnault et al.

Loven J, Orlando DA, Sigova AA, Lin CY, Rahl PB, Burge CB, Valouev A, Johnson DS, Sundquist A, Medina C, Anton E, Levens DL, Lee TI, Young RA. 2012. Revisiting global gene Batzoglou S, Myers RM, Sidow A. 2008. Genome-wide expression analysis. Cell 151: 476–482. analysis of transcription factor binding sites based on ChIP- Medjkane S, Perez-Sanchez C, Gaggioli C, Sahai E, Treisman R. Seq data. Nat Methods 5: 829–834. 2009. Myocardin-related transcription factors and SRF are Vartiainen MK, Guettler S, Larijani B, Treisman R. 2007. Nuclear required for cytoskeletal dynamics and experimental metas- actin regulates dynamic subcellular localization and activity tasis. Nat Cell Biol 11: 257–268. of the SRF cofactor MAL. Science 316: 1749–1752. Mikkelsen TS, Xu Z, Zhang X, Wang L, Gimble JM, Lander ES, Wang D, Chang PS, Wang Z, Sutherland L, Richardson JA, Small Rosen ED. 2010. Comparative epigenomic analysis of mu- E, Krieg PA, Olson EN. 2001. Activation of cardiac gene rine and human adipogenesis. Cell 143: 156–169. expression by myocardin, a transcriptional cofactor for Miralles F, Posern G, Zaromytidou AI, Treisman R. 2003. Actin serum response factor. Cell 105: 851–862. dynamics control SRF activity by regulation of its coactiva- Wang Z, Wang DZ, Hockemeyer D, McAnally J, Nordheim A, tor MAL. Cell 113: 329–342. Olson EN. 2004. Myocardin and ternary complex factors Mokalled MH, Johnson A, Kim Y, Oh J, Olson EN. 2010. compete for SRF to control smooth muscle gene expression. Myocardin-related transcription factors regulate the Cdk5/ Nature 428: 185–189. Pctaire1 kinase cascade to control neurite outgrowth, neu- Yu FX, Zhao B, Panupinthu N, Jewell JL, Lian I, Wang LH, Zhao ronal migration and brain development. Development 137: J, Yuan H, Tumaneng K, Li H, et al. 2012. Regulation of the 2365–2374. Hippo-YAP pathway by G-protein-coupled receptor signal- Morita T, Mayanagi T, Sobue K. 2007. Dual roles of myocardin- ing. Cell 150: 780–791. related transcription factors in epithelial mesenchymal tran- Zaromytidou AI, Miralles F, Treisman R. 2006. MAL and sition via slug induction and actin remodeling. J Cell Biol ternary complex factor use different mechanisms to contact 179: 1027–1042. a common surface on the serum response factor DNA- Obrietan K, Impey S, Storm DR. 1998. Light and circadian binding domain. Mol Cell Biol 26: 4134–4148. rhythmicity regulate MAP kinase activation in the supra- Zhang Y, Liu T, Meyer CA, Eeckhoute J, Johnson DS, Bernstein chiasmatic nuclei. Nat Neurosci 1: 693–700. BE, Nusbaum C, Myers RM, Brown M, Li W, et al. 2008. Olson EN, Nordheim A. 2010. Linking actin dynamics and gene Model-based analysis of ChIP-seq (MACS). Genome Biol 9: transcription to drive cellular motile functions. Nat Rev Mol R137. Cell Biol 11: 353–365. Paszek MJ, Zahir N, Johnson KR, Lakins JN, Rozenberg GI, Gefen A, Reinhart-King CA, Margulies SS, Dembo M, Boettiger D, et al. 2005. Tensional homeostasis and the malignant phenotype. Cancer Cell 8: 241–254. Posern G, Treisman R. 2006. Actin’ together: serum response factor, its cofactors and the link to signal transduction. Trends Cell Biol 16: 588–596. Provenzano PP, Inman DR, Eliceiri KW, Beggs HE, Keely PJ. 2008. Mammary epithelial-specific disruption of focal adhesion kinase retards tumor formation and metastasis in a transgenic mouse model of human breast cancer. Am J Pathol 173: 1551–1565. Provenzano PP, Inman DR, Eliceiri KW, Keely PJ. 2009. Matrix density-induced mechanoregulation of breast cell phenotype, signaling and gene expression through a FAK-ERK linkage. Oncogene 28: 4326–4343. Qiu P, Ritchie RP, Fu Z, Cao D, Cumming J, Miano JM, Wang DZ, Li HJ, Li L. 2005. Myocardin enhances Smad3-mediated transforming growth factor-b1 signaling in a CArG box- independent manner: Smad-binding element is an important cis element for SM22a transcription in vivo. Circ Res 97: 983–991. Sordella R, Jiang W, Chen GC, Curto M, Settleman J. 2003. Modulation of Rho GTPase signaling regulates a switch between adipogenesis and myogenesis. Cell 113: 147–158. Sullivan AL, Benner C, Heinz S, Huang W, Xie L, Miano JM, Glass CK. 2011. Serum response factor utilizes distinct promoter- and enhancer-based mechanisms to regulate cytoskeletal gene expression in macrophages. Mol Cell Biol 31: 861–875. Swift J, Ivanovska IL, Buxboim A, Harada T, Dingal PC, Pinter J, Pajerowski JD, Spinler KR, Shin JW, Tewari M, et al. 2013. Nuclear lamin-A scales with tissue stiffness and enhances matrix-directed differentiation. Science 341: 1240104. Thurman RE, Rynes E, Humbert R, Vierstra J, Maurano MT, Haugen E, Sheffield NC, Stergachis AB, Wang H, Vernot B, et al. 2012. The accessible chromatin landscape of the human genome. Nature 489: 75–82.

16 GENES & DEVELOPMENT Downloaded from genesdev.cshlp.org on October 7, 2021 - Published by Cold Spring Harbor Laboratory Press

Rho-actin signaling to the MRTF coactivators dominates the immediate transcriptional response to serum in fibroblasts

Cyril Esnault, Aengus Stewart, Francesco Gualdrini, et al.

Genes Dev. published online April 14, 2014 Access the most recent version at doi:10.1101/gad.239327.114

Supplemental http://genesdev.cshlp.org/content/suppl/2014/04/07/gad.239327.114.DC1 Material

Related Content Dual views of SRF: a genomic exposure Kathleen A. Clark and Barbara J. Graves Genes Dev. May , 2014 28: 926-928

Published online April 14, 2014 in advance of the full issue.

Creative This article, published in Genes & Development, is available under a Creative Commons Commons License (Attribution 4.0 International), as described at License http://creativecommons.org/licenses/by/4.0.

Email Alerting Receive free email alerts when new articles cite this article - sign up in the box at the top Service right corner of the article or click here.